WHAT THIS DOES Drop a bank statement PDF into one Google Drive folder. Gemini reads the pages — including scanned ones, where PDF text extraction returns nothing — and every deposit lands as a row in your Google Sheet. Nothing leaves Google. No third-party PDF service. 1. READ THE STATEMENT Gemini takes the PDF directly: up to 50MB or 1000 pages, no conversion step. Set your own Gemini credential (Google AI Studio, free tier, no card). If the next node sees nothing, open this node's output and point its Text field at whatever key holds the markdown. 2. PULL OUT THE DEPOSITS The Information Extractor reads the markdown and returns one object per deposit: date, description, amount. Edit the schema here to pull withdrawals or a balance instead. 3. YOUR TABLE Split Out turns the array into one item per row, and Sheets appends them. Replace both REPLACE_WITH_ ids with your own folder and sheet. THE VIDEO https://youtube.com/shorts/RvonZo-QTRY IMPORT 1. Download the JSON attached to this post. 2. In n8n: Workflows → the ⋯ menu → Import from File. 3. Wire it as above, then toggle it Active.

AGENT TOOLS FOR RAG TOOL TO ADD A GOOGLE DRIVE FILE TO VECTOR DB RAG AI AGENT WITH CHAT INTERFACE ASK YOUR OWN PDFS — GOOGLE DRIVE → SUPABASE → CHAT (GEMINI, FREE TIER) WHAT IT DOES Drop a PDF (or a Google Doc) into one Drive folder. n8n downloads it, pulls the text out, cuts it into chunks, turns each chunk into a Gemini embedding and stores it in Supabase. Then open the chat, ask a question in plain words, and the agent reads your own documents to answer. Update a file and its old chunks are replaced. YOU NEED (ALL ON FREE PLANS) n8n · a Google account (Drive) · a Supabase project (free tier) · a Gemini key from Google AI Studio (free tier, no card). BEFORE WIRING — RUN THIS ONCE IN SUPABASE → SQL EDITOR create extension if not exists vector; create table documents (id bigserial primary key, content text, metadata jsonb, embedding vector(3072)); create or replace function match_documents (query_embedding vector(3072), match_count int default null, filter jsonb default '{}') returns table (id bigint, content text, metadata jsonb, similarity float) language plpgsql as $$ begin return query select id, content, metadata, 1 - (documents.embedding <=> query_embedding) as similarity from documents where metadata @> filter order by documents.embedding <=> query_embedding limit match_count; end; $$; (3072 is the size of a gemini-embedding-001 vector. If you pick a different embedding model, change both numbers to its size.) WIRE IT (3 CREDENTIALS, 10 NODES) 1. File Created and File Updated → Credential → Google Drive OAuth2 → sign in, then pick the folder to watch in both. 2. Download File → the same Google Drive credential. 3. Delete Old Doc Rows → Supabase API → your project URL and the service_role key from Supabase → Project Settings → API. 4. Insert into Supabase and Supabase Vector Store → the same Supabase credential; table is documents, query name is match_documents. 5. Gemini Embeddings, Gemini Embeddings (insert), Gemini Chat Model and Gemini Chat Model (tool) → Google Gemini (PaLM) API → paste your key. Keep the embedding model the SAME in both embedding nodes.

TELEGRAM VOICE ASSISTANT → GOOGLE CALENDAR + GMAIL (GEMINI, FREE TIER) WHAT IT DOES Send the bot a voice note or a text. A voice note is downloaded and Gemini turns it into words. The AI Agent reads the request and uses five tools: read your calendar, read your unread mail, look a person up in a contacts sheet, create a calendar event, send an email. It replies on Telegram with exactly what it did. YOU NEED (ALL ON FREE PLANS) n8n · a Telegram bot token from @BotFather · a Google account (Calendar, Gmail, Sheets) · a Gemini key from Google AI Studio (free tier, no card). BEFORE WIRING Make a Google Sheet called Contacts with two columns, name and email, and add the people you talk to. WIRE IT (4 CREDENTIALS, 9 NODES) 1. Telegram Trigger, Get Voice File, Reply on Telegram → Credential → Telegram API → paste the bot token. 2. Transcribe with Gemini and Gemini Chat Model → Google Gemini (PaLM) API → paste your key. Pick a Flash model in both if the listed one is gone. 3. Read Calendar and Create Event → Google Calendar OAuth2 → sign in, then pick your calendar in both nodes. 4. Read Inbox and Send Email → Gmail OAuth2 → same sign-in. 5. Contacts → Google Sheets OAuth2 → same sign-in, then pick the Contacts sheet and its tab. 6. Turn the workflow on. Open your bot in Telegram, send a voice note: "book half an hour with Sara tomorrow and send her the invite". TIP The Memory node keeps the last messages per chat, so "make it an hour" changes the meeting you just booked. Adapted from Derek Cheung's community template "Angie, personal AI assistant" — model swapped to Gemini, Baserow tools replaced by a Google Sheet, and the two write tools (Create Event, Send Email) added. VOICE OR TEXT A text goes straight to the agent. A voice note takes the top path: download, then Gemini turns it into words. IMPORT 1. Download the JSON attached to this post. 2. In n8n: Workflows → the ⋯ menu → Import from File. 3. Wire it as above, then toggle it Active. THE VIDEO

INVOICE EMAIL → DRIVE → GEMINI → AIRTABLE ROWS WHAT IT DOES Every minute the Gmail Trigger looks for new mail with an attachment. Each PDF is saved to a Drive folder, the text is pulled out of the PDF, Gemini reads it and returns the company name, the total and every line item as JSON. One Airtable row per invoice (subject, sender, Drive link, total) and one row per line item, linked to it. Nobody types an invoice. YOU NEED (ALL FREE) n8n · a Google account (Gmail + Drive) · a Gemini key from Google AI Studio (free tier, no card) · an Airtable account (free plan). AIRTABLE: MAKE TWO TABLES IN ONE BASE 1. Table 1 (invoices): fields email_subject (text), email (text), invoice_link (URL), amount (number). 2. invoice_line_items: fields line_item (text), amount (number), invoice (link to Table 1). WIRE IT (4 CREDENTIALS, 6 NODES) 1. Gmail Trigger and Gmail → Credential → Gmail OAuth2 → sign in. 2. Upload to Google Drive and Download Binary Data → Google Drive OAuth2 → same sign-in. On Upload to Google Drive pick the folder the PDFs go to. 3. Google Gemini Chat Model → Google Gemini (PaLM) API → paste your key. 4. Add invoice and Airtable1 → Airtable Personal Access Token → then pick your base and the two tables. The column mapping is already filled in; it matches the field names above. 5. Toggle the workflow Active. Send yourself an email with a PDF invoice attached and watch the rows appear. MAKE IT YOURS (1 EDIT) Invoice Data Extractor → the prompt lists the four things it pulls out. Add a line (due date, invoice number, VAT) and add the same field to the Structured Output Parser schema and to the Add invoice mapping. NOTE The Filter after the trigger drops mail with no attachment. Every attachment on a matching email is treated as an invoice, so give this workflow its own label or inbox if your mail carries other PDFs. THE VIDEO https://youtube.com/shorts/pZ9FiD2iw3U IMPORT 1. Download the JSON attached to this post.